Fraud Detection in Automotive Insurance Domain Using Machine Learning

Abstract

The insurance industry is one of the most vital aspects of modern society, the economy, and people's lives. It brings peace and security to people by offsetting the financial risks of damage and loss. Fraud in the Automotive Insurance domain illustrates a scenario in which the client obtains money by providing falsified papers, demonstrating fake accidents or economic claims for early losses. One of the biggest harmful challenges here is an interaction between policyholders and the insurance industry which creates a possible situation for fraud claims. By keeping these in mind, in this research, fraud detection in the automotive insurance domain (AID) is performed using machine learning (ML) techniques. It is a binary classification problem that utilized the recursive feature elimination (RFE) method to uncover the most influential factors for detecting fraud. The significance of each attribute is also determined. Moreover, a comparative analysis of the ML models is performed. The experimental outcomes show that the multilayer perceptron model is quite robust and outperforms the other algorithms in detecting fraud.